---
title: "NEON4CAST Dashboard"
output:
  flexdashboard::flex_dashboard:
    theme: 
      version: 4
      bootswatch: lux
    orientation: columns
    vertical_layout: fill
    source_code: embed
---

```{r setup, include=FALSE}
library(flexdashboard)
library(tidyverse)
library(plotly)
library(ggiraph)
library(clock)
library(dbplyr)
library(RSQLite)
source("R/plotly_helpers.R")

thematic::thematic_rmd(font = "auto")
```


Home
=====



```{r include=FALSE}
combined <- read_csv("https://data.ecoforecast.org/analysis/combined_forecasts_scores.csv.gz")
```


Column {data-width=650}
-----------------------------------------------------------------------


### NEON Ecological Forecasting Challenge sites

```{r}
## FIXME color code by number of challenges at each site?

challenges <- combined %>% select(theme, siteID) %>% distinct() %>%
  separate(siteID, into = c("siteID", "plot")) %>%
  select(theme, siteID) %>% 
  distinct() 
  
library(sf)
library(tmap)
geo <- jsonlite::read_json("https://github.com/eco4cast/neon4cast/raw/main/inst/extdata/geo.json", TRUE)
site_id <- gsub(", .*$", "", geo$geographicDescription)
bb <- geo$boundingCoordinates[1:4] %>% mutate_all(as.numeric) %>% mutate(siteID = site_id)
bb <- left_join(bb, challenges, by = "siteID")
neon <- st_as_sf(bb, coords = c("westBoundingCoordinate", "northBoundingCoordinate"), crs = 4326)

tmap::tmap_mode("view")
tm_shape(neon) + tm_dots(col="theme", alpha=.4, size = .1)
```

Column {data-width=350}
-----------------------------------------------------------------------

## Stats

### Challenges 


```{r}
flexdashboard::valueBox(5, color = "primary")
```

### Teams

```{r}
total_teams <- combined %>% select(team) %>% distinct() %>% count()
flexdashboard::valueBox(total_teams, color = "success")
```



### Total Forecasts

```{r}
total_forecasts <- combined %>% select(team, forecast_start_time) %>% distinct() %>% count()
flexdashboard::valueBox(total_forecasts, color = "info")
```





Phenology
==========


Column {data-width=650}
-----------------------------------------------------------------------

### Phenology (Greeness)

```{r}
## determine these more cleverly
start <- as.Date("2021-05-01")
end <- Sys.Date() %>% clock::add_months(1)

## Get most recent submission per team
pheno_teams <- combined %>% filter(theme == "phenology") %>%
  select(team, forecast_start_time) %>% distinct() %>%
  group_by(team) %>%
  slice_max(forecast_start_time)

pheno_latest <- inner_join(pheno_teams, combined)

p <- pheno_latest %>% 
  filter(time > start, time < end, target == "gcc_90") %>%
  ggplot() +
  geom_ribbon(aes(x = time, ymin = lower95, ymax = upper95, fill = team), alpha = 0.2) +
  geom_line(aes(time, mean, col = team)) +
  geom_point(aes(time, obs), size = .1) + 
  facet_wrap(~siteID)

gp <- plotly::ggplotly(p)
gp <- patch_legend(gp)

gp
```
### Phenology (Redness)

```{r}
p <- pheno_latest %>% 
  filter(time > start, time < end, target == "rcc_90") %>%
  ggplot() +
  geom_ribbon(aes(x = time, ymin = lower95, ymax = upper95, fill = team), alpha = 0.2) +
  geom_line(aes(time, mean, col = team)) +
  geom_point(aes(time, obs), size = .1) + 
  facet_wrap(~siteID)

gp <- plotly::ggplotly(p)
gp <- patch_legend(gp)

gp
```
Column {data-width=350}
-----------------------------------------------------------------------


### Teams

```{r}
total <- combined %>% filter(theme == "phenology") %>% select(team) %>% distinct() %>% count()
flexdashboard::valueBox(total)
```

### Leaderboard (target: greeness)

```{r}
con <- DBI::dbConnect(RSQLite::SQLite(), tempfile())
dbWriteTable(con, "combined", combined, overwrite=TRUE)

pheno <- tbl(con, "combined") %>% filter(theme == "phenology") %>%
  select(siteID, target, time, team, forecast_start_time, crps)

## expand a table to all possible observations (target, siteID, time)
## for each team, for each forecast_start_time:
all <- pheno %>% expand(team, target, siteID, time, forecast_start_time)

## Use this list to make explicit NA for any observation for which a forecast was not provided
na_filled <- pheno %>% right_join(all)

## Fill in any missing observation with the most recent forecast made prior to the start_time
self_fill <- na_filled %>%
  window_order(team, target, time, forecast_start_time) %>% # 
  group_by(team, target, siteID, time) %>% 
  fill(crps, .direction="up")

## We will now fill all remaining NAs using the NULL forecast:
null_score <- self_fill %>% ungroup() %>%
  filter(team == "EFInull") %>% rename(null = crps) %>% select(-team)

all_filled <- self_fill %>%
  left_join(null_score) %>% # add null-score as a separate column
  mutate(filled_crps = case_when(is.na(crps) ~ null,
                                 !is.na(crps) ~ crps))


scores <- all_filled %>% filter(target == "gcc_90") %>%
  group_by(team) %>% 
  summarise(mean_crps = mean(filled_crps)) %>%
  collect() %>% arrange(mean_crps)

scores %>%
  rmarkdown::paged_table()
```

### Leaderboard (target: redness)

```{r}

scores <- all_filled %>% filter(target == "rcc_90") %>%
  group_by(team) %>% 
  summarise(mean_crps = mean(filled_crps)) %>%
  collect() %>% arrange(mean_crps)

scores %>%
  rmarkdown::paged_table()
```

### Days remaining

```{r}
#days <- (pheno_end_date - Sys.Date() ) 
#max <- pheno_end_date - pheno_start_date
#gauge(days, min = 0, max = max, symbol = '', gaugeSectors(
#  success = c(81, max), warning = c(10, 3), danger = c(0, 2)
#))
```



Aquatics
========

Column {data-width=650}
-----------------------------------------------------------------------


### Aquatics Forecasts

```{r}

start <- as.Date("2021-05-31")
end <- as.Date("2021-08-31")

## Get most recent submission per team
aq_teams <- combined %>% filter(theme == "aquatics") %>%
  select(team, forecast_start_time) %>% distinct() %>%
  group_by(team) %>%
  slice_max(forecast_start_time)

## Heck show all the forecasts
p <- combined %>% #inner_join(aq_teams) %>%
  filter(theme == "aquatics", time >= start, time <= end) %>%
  ggplot() +
  geom_ribbon(aes(x = time, ymin = lower95, ymax = upper95, fill = team), alpha = 0.2) +
  geom_point(aes(time, obs)) + 
  geom_line(aes(time, mean, col = team)) +
  facet_grid(target~siteID, scales = "free")

gp <- plotly::ggplotly(p)
gp <- patch_legend(gp)

gp
```

Column {data-width=350}
-----------------------------------------------------------------------

### Teams

```{r}
total <- combined %>% filter(theme == "aquatics") %>% select(team) %>% distinct() %>% count()
flexdashboard::valueBox(total)
```

### Leaderboard

```{r}
combined %>% 
  filter(theme == "aquatics") %>%
  group_by(team) %>%
  summarise(mean_crps = mean(crps,na.rm=TRUE)) %>%
  arrange(mean_crps) %>%
  rmarkdown::paged_table()
```

### Days elapsed

```{r}
days <- end-start
gauge(days, min = 0, max = end-start, symbol = '', gaugeSectors(
  success = c(11, as.numeric(end-start)), warning = c(10, 3), danger = c(0, 2)
))
```



Terrestrial
===========

Column {data-width=650}
-----------------------------------------------------------------------

### Terrestrial Forecasts (Daily)

```{r}
## Could consider displaying older ones
start <- combined %>% 
  filter(theme == "terrestrial_daily") %>%
  select(forecast_start_time) %>% 
  distinct() %>% 
  arrange(desc(forecast_start_time))

p <- combined %>%
  filter(theme == "terrestrial_daily", forecast_start_time == start[[2,1]]) %>%
  ggplot() +
  geom_ribbon(aes(x = time, ymin = lower95, ymax = upper95, fill = team), alpha = 0.2) +
  geom_line(aes(time, mean, col = team)) +
  geom_point(aes(time, obs)) + 
  facet_grid(target ~ siteID, scales = "free")

gp <- plotly::ggplotly(p)
gp <- patch_legend(gp)

gp
```


### Terrestrial Forecasts (30 minute)

```{r}
## Could consider displaying older ones
start <- combined %>% 
  filter(theme == "terrestrial_30min") %>%
  select(forecast_start_time) %>% 
  distinct() %>% 
  arrange(desc(forecast_start_time))

p <- combined %>%
  filter(theme == "terrestrial_30min", forecast_start_time == start[[2,1]]) %>%
  ggplot() +
  geom_ribbon(aes(x = time, ymin = lower95, ymax = upper95, fill = team), alpha = 0.2) +
  geom_line(aes(time, mean, col = team)) +
  geom_point(aes(time, obs)) + 
  facet_grid(target ~ siteID, scales = "free")

gp <- plotly::ggplotly(p)
gp <- patch_legend(gp)

gp
```

Column {data-width=350}
-----------------------------------------------------------------------


### Teams: `terrestrial_daily`

```{r}
total <- combined %>% filter(theme == "terrestrial_daily") %>%
  select(team) %>% distinct() %>% count()
flexdashboard::valueBox(total)
```

### Teams: `terrestrial_30min`

```{r}
total <- combined %>% filter(theme == "terrestrial_30min") %>% 
  select(team) %>% distinct() %>% count()
flexdashboard::valueBox(total)
```



### Leaderboard (daily)

```{r}
combined %>% 
  filter(theme == "terrestrial_daily") %>%
  group_by(team) %>%
  summarise(mean_crps = mean(crps,na.rm=TRUE)) %>%
  arrange(mean_crps) %>%
  rmarkdown::paged_table()
```
### Leaderboard (30 minute)

```{r}
combined %>% 
  filter(theme == "terrestrial_30min") %>%
  group_by(team) %>%
  summarise(mean_crps = mean(crps,na.rm=TRUE)) %>%
  arrange(mean_crps) %>%
  rmarkdown::paged_table()
```

Ticks
=======

Column {data-width=650}
-----------------------------------------------------------------------




### Ticks

```{r}
## Could consider displaying older ones
start <- combined %>% 
  filter(theme == "ticks") %>%
  select(forecast_start_time) %>% 
  distinct() %>% 
  arrange(desc(forecast_start_time))

p <- combined %>%
  filter(theme == "ticks", forecast_start_time == start[[2,1]]) %>% # second most recent start time
  ggplot() +
  geom_ribbon(aes(x = time, ymin = lower95, ymax = upper95, 
                  fill = team, lty=target), alpha = 0.2) +
  geom_line(aes(time, mean, col = team, lty=target)) +
  geom_point(aes(time, obs, shape=target)) + 
  facet_wrap(~siteID)


## ggiraph also supports ggplot-syntax-based controls
ggiraph(ggobj = p)


```


Column {data-width=350}
-----------------------------------------------------------------------


### Teams

```{r}
total <- combined %>% filter(theme == "ticks") %>% 
  select(team) %>% distinct() %>% count()
flexdashboard::valueBox(total)
```



### Leaderboard

```{r}
combined %>% 
  filter(theme == "ticks") %>%
  group_by(team) %>%
  summarise(mean_crps = mean(crps,na.rm=TRUE)) %>%
  arrange(mean_crps) %>%
  rmarkdown::paged_table()
```


Beetles
=======

Column {data-width=650}
-----------------------------------------------------------------------

### Beetles Forecasts

```{r fig.width=8, fig.height=16}
## determine these more cleverly
start <- combined %>% 
  filter(theme == "beetles") %>%
  select(forecast_start_time) %>% 
  distinct() %>% 
  arrange(desc(forecast_start_time))

p <- combined %>%
  filter(theme == "beetles", 
         target == "richness",
         forecast_start_time == start[[1,1]]) %>% # second most recent start time
  ggplot() +
  geom_ribbon(aes(x = time, ymin = lower95, ymax = upper95, fill = team), alpha = 0.2) +
  geom_line(aes(time, mean, col = team)) +
  geom_point(aes(time, obs)) + 
  facet_wrap(~siteID)

ggiraph(ggobj = p)


```


Column {data-width=350}
-----------------------------------------------------------------------


### Teams

```{r}
total <- combined %>% filter(theme == "ticks") %>% 
  select(team) %>% distinct() %>% count()
flexdashboard::valueBox(total)
```


### Leaderboard

```{r}
combined %>% 
  filter(theme == "beetles") %>%
  group_by(team) %>%
  summarise(mean_crps = mean(crps,na.rm=TRUE)) %>%
  arrange(mean_crps) %>%
  rmarkdown::paged_table()
```